Hi! are there known issues with Gemini 3+BAML? I'm getting weird errors where it looks like generation is stopped right away on code/prompts that work with Gemini 2.5/claude 4.5:

2025-11-19T10:02:52.660 [BAML WARN] Function ValidateTest:
    Client: VertexGeminiPro3 (gemini-3-pro-preview) - 62697ms. StopReason: STOP. Tokens(in/out): 719/unknown
[...]
Test AE42_STD_FTOO generated an exception: BamlValidationError(message=Failed to parse LLM response: Failed to coerce value: <root>: Failed while parsing required fields: missing=2, unparsed=0

The same prompt works in the Google cloud console Might still be caused by issues on google's end ? Something like 15% of my requests fail like this with Gemini 3+Baml By the way it was great to meet you at DotAI Vaibhav! I'm proudly rocking the Baaaaml sticker on my corporate laptop 🫡

can you by chance share your schema + 1 test case?

yup, sending it to you via dm 🙂

FWIW, I'm using gemini 3 pro preview with BAML & works great.

are you using Vertex or the AI studio api ?

FWIW I’ve also been using Gemini 3 pro preview using both vertex and AI studio it works fine for me. But, I’m calling it via open router and not google directly.

Only 15%? That's the puzzling part... Does OpenRouter tells you if you get rate limited?

Direct API